0

My dad recently recorded a 3h long meeting with a smartphone app (I don't know the name, but I can ask him) and this app records in .m4a files.

The only significant sound in the ambiance were human voices.

I don't know why, but this recording came out with a strong clicking sound that makes it almost impossible to comprehend anything.

The clicking happens together with the audio, the louder the voice, the louder the clicks.

I investigated the waveform and found what seems to be some weird amplitude modulation happening:

enter image description here

This is what the signal looks like up close, this stretch has around 0.15 s of duration.

And those apparently silent parts actually have signal, just very low, but I can amplify.

enter image description here

Do you have any idea of what might have caused that and how I could make it a bit better?

So far the only thing I could think about is to find the pattern of this modulation and apply an inverse modulation, but I'm having trouble figuring out just what this pattern is, IF it has a constant pattern.

If the modulation pattern varies along the signal this is going to be way more complicated.

Thanks!

1
  • 3
    We can help a lot more if we can hear the audio. Seeing the waveform only is quite unhelpful. May 20, 2023 at 17:46

2 Answers 2

0

My best guess would be that the codec for the m4a format does not match your decoder (one of them might be defective). Your best bet may be to let your father's smartphone (preferably using the originating app) convert the .m4a file to a .wav file and see whether that ends up cleaner on your computer than what you got now.

0

you could use RX to declick your audio. RX is somewhat costly but if that isn't an issue then it's potentially your best shot since you said that there's amplitude modulation

https://www.izotope.com/en/products/rx/features/de-click.html

there are various algorithms within the tool that are frequency specific so that it will adaptively de-click your audio

the following list is from rx8 which is two generations older than the current, and seems like you could make use of multi-band or low latency algorithms:

https://s3.amazonaws.com/izotopedownloads/docs/rx8/en/de-click/index.html

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.